The YouTube Algorithm in 2026: How It Picks Videos
The YouTube algorithm, in one sentence: it is an expanding test. A new video is shown to a small audience first, and if enough of those people click and keep watching, it goes to a larger one, then larger again. This page explains that model in plain terms and what it means for your script, your title, and the length you choose in 2026.
What follows is Prepublish's explanation of publicly described behavior, not a YouTube document: the ranking factors themselves are not published, so treat this as a working model rather than a spec.
Creators love to blame the algorithm. "The algorithm is suppressing my content." "The algorithm changed and now my videos don't get views." "I need to figure out the algorithm."
Here is the thing nobody wants to hear: the algorithm is not complicated. It is not a mystery. It is not out to get you.
It does one thing. It finds videos that people watch for a long time, and it shows those videos to more people. That is it. Everything else is noise.
If your videos hold attention, the algorithm will find your audience for you. If they do not, no amount of SEO tricks, posting schedules, or hashtag strategies will save you.
How the Algorithm Actually Decides
Forget everything you have read about "hacking" the algorithm. Here is what actually happens when you upload a video.
YouTube shows your video to a small group of people. Maybe a few hundred. It watches what they do. Whether they click matters, but whether they stay matters more.
If that first group watches a large percentage of your video, YouTube thinks "this might be good" and shows it to a slightly larger group. Same test. If the second group also watches, it expands again. And again.
This is the entire algorithm. It is an expanding test. Your video either passes each round by holding attention, or it fails and YouTube stops expanding.
The algorithm does not care about your upload schedule. It does not care about your tags. It does not care how many videos you have published. It cares about one thing: when we show this video to people, do they watch it?
The Two Numbers That Matter
Out of every metric YouTube gives you, only two actually drive the algorithm's decisions.
Click-through rate (CTR). What percentage of people who see your thumbnail actually click on it. This is entirely controlled by your title and thumbnail. A great CTR is above 8%. Most creators sit between 3-6%.
Average view duration (AVD). Of the people who click, how long do they watch? This is controlled by your script and editing. A video with 50% AVD (people watch half of it on average) will massively outperform a video with 30% AVD, even if the second video has a higher CTR.
YouTube multiplies these two numbers together to estimate what they call "expected watch time per impression." The videos with the highest expected watch time get the most impressions. That is the algorithm in one sentence.
This means you need both. A great thumbnail with a boring script gives you high CTR but low AVD. The algorithm tests it, sees people leaving quickly, and stops showing it. A mediocre thumbnail with an incredible script means nobody clicks in the first place.
The sweet spot is a thumbnail and title that accurately represent content that holds attention. Not clickbait. Not bait and switch. An honest promise, delivered well.
What the Algorithm Is NOT Doing
Let me clear up some myths that waste creators' time.
The algorithm does not punish you for not posting. If you skip a week, your next video will be tested the same way as any other video. The test is per-video, not per-channel. Your upload frequency has zero direct impact on how any individual video performs.
The algorithm does not have a "niche box." YouTube does not categorize your channel and only show your videos to people who watch that category. It shows your video to whoever is most likely to watch it, based on their individual viewing history. If you make a cooking video and a gaming video, each one reaches different audiences. Neither one hurts the other.
The algorithm does not suppress small channels. Small channels have less data, so YouTube is more cautious with testing. But a video from a 500-subscriber channel that holds its audience well has a real chance of being pushed. The algorithm does not check your subscriber count before deciding to recommend you.
The algorithm does not care about keywords in your description. YouTube's AI understands what your video is about by watching and listening to it. Stuffing keywords into your description is a 2015 strategy that no longer matters.
Why Most Algorithm Advice Is Wrong
The YouTube advice industry has a problem. Most of it is based on correlation, not causation.
"Creators who post 3 times a week grow faster." Sure. But that is because creators who post 3 times a week get more practice writing scripts and making thumbnails. It is not because the algorithm rewards a posting schedule.
"Longer videos rank better." They do not. Longer videos accumulate more total watch time per view (because there are more minutes to watch), but the algorithm cares about the percentage of the video watched, not the total minutes. A 5-minute video that holds most of its audience gives the algorithm a stronger signal than a 20-minute video that loses most of its audience early.
"You need to post within a certain time window." Your posting time affects who sees it first, which can impact the initial test group. But if the video is good, it will get pushed regardless of when you posted. Optimizing posting time is a 2% improvement at best. Your script is a 200% improvement.
The Only Algorithm Strategy That Works
Stop trying to game the system. Instead, work with it.
The algorithm wants to find good videos. Make good videos. Specifically:
Write scripts that hold attention from the first sentence. Test your title and thumbnail to make sure people click. Deliver on the promise your title makes. End before the viewer wants you to.
Do those four things consistently, and the algorithm becomes your best friend. It will find your audience, grow your reach, and push your content to people who have never heard of you.
The algorithm is not the problem. It never was. The problem is what happens after someone clicks your video. Fix that, and everything else follows.
Try it on your own script
Paste your draft below. You get your hook, structure, and pacing scores, a script-level attention-risk map, and the single biggest issue quoted from your own lines. Free, no login.
Free · No login · See a sample audit first if you prefer.
Frequently asked questions
How does the YouTube algorithm decide which videos to promote?
It runs an expanding test. Your video is shown to a small group, and if they click and keep watching, it goes to a larger group, then larger again. The two numbers driving each round are click-through rate (controlled by title and thumbnail) and average view duration (controlled by script and editing). YouTube multiplies them into expected watch time per impression, and the highest scores get the most impressions.
Does upload frequency affect the YouTube algorithm?
Not directly. The algorithm tests each video on its own performance, so skipping a week does not penalize your next upload. Creators who post more often tend to grow faster because they practice more and produce more chances to win, not because a schedule is rewarded. One strong video beats three rushed ones.
Does YouTube suppress small channels?
No. Small channels have less watch history for YouTube to learn from, so initial testing is more cautious, but a 500-subscriber channel whose videos hold their audience still gets tested. The algorithm does not check subscriber count before recommending a video; it checks whether the people it tested the video on kept watching.
What does the YouTube algorithm optimize for?
In the model on this page, it optimizes the chance that a viewer clicks and then keeps watching, which YouTube surfaces as expected watch time per impression. Click-through rate comes from the title and thumbnail; how long people stay comes from the script and the edit. This is our explanation of publicly described behavior, not a published YouTube specification.
Did the YouTube algorithm change in 2026?
We do not cite a primary source for a 2026 change on this page, so we are not going to claim one. The mechanism described here, a small test audience that expands when people keep watching, is the model we have used, and the practical advice that follows from it is unchanged: make the title match the opening, and give the viewer a reason to stay past it. For product changes themselves, read YouTube's own announcements.
Related Articles
The Creator Who Got 500K Views and Gained Nothing
A viral video sounds great. But if nobody watches past 30 seconds, YouTube stops recommending you. Views without retention is a trap creators fall into.
Your Videos Are Bleeding Viewers. Here Is Where and Why.
You are losing viewers at specific moments in every video you publish. Not because the content is bad, but because your structure has invisible leaks.
Twelve Script Patterns to Check Before You Record
A craft checklist for the script decisions that shape attention: where value lands, how transitions read, how sentences vary, and where a draft goes quiet.
Free tools to put this into practice
Want to analyze your own scripts?
Start Script Analysis